The influence of the seed size, loadings and the batch time on crystallization kinetics
and granulometric properties of the obtained pentaerythritol crystals has been
investigated. Experiments have been performed in the laboratory batch cooling
crystallizer (Rushton dimensions). The solubility curve and the metastable zone width
have been determined at the defined cooling profile. In order to compare the obtained
results, an experiment of unseeded crystallization was performed. The experiments have
been performed within the metastable zone width.
For estimation of the growth kinetics, from the desupersaturation curve during batch
seeded experiments, the Garside et al. method was used. The crystal growth is
controlled by the surface integration (polynuclear or birth and spread mechanism of
grow), since the values of the exponent of growth, g, are very high (g>2).
The crystal size distribution can be approximated with the RRSB function. The smaller
seed size and lower seed loadings leads to a uni-modal size distribution of the final
crystals and the higher crystal growth rate. Decreasing the batch time from 80 min to 35
min, improves the final crystal size distribution. (unimodal for lower seed loadings).
1. Introduction
Crystallization is one of the most complex separation processes due to numerous
simultaneous processes (momentum, heat and mass transfer, phase change, dissolution,
agglomeration, breakage,...) (Myerson, 2002). The obtained crystals are of the specific
purity and granulometric characteristics (shape, size, and crystal size distribution). The
crystal quality is very sensitive to cooling rate, mixing intensity, system geometry, type
of crystallization process and impeller used for mixing. Crystals of the desirable
granulometric properties can be obtained only if the thermal and the hydrodynamic
conditions are selected very carefully (Gates, 2003). In order to minimize excessive
primary nucleation, crystallization has to be conducted within the metastable zone
width. Excessive nucleation yields too many small crystals and an unacceptable crystal
size distribution. For the same reason, secondary nucleation (contact nucleation,
breakage, mixing) has to be suppressed as much as possible (Gates, 2003). At the
optimal process conditions, regular shape crystals with the unimodal narrow size
distribution will be produced (Chong et al, 2004). Hydrodinamic conditions have to
provide good flow of the solution past the heat transfer surfaces and satisfactory
handling of the crystals being formed (Mersmann, 1988). In other words, to support
crystal growth, it is very important that crystal surface area is accessible at locations
where the supersaturation is generated.
1.1 Seeded crystallization
The crystallization process (nucleation and growth) can be controlled by adding a
certain number of seed crystals of a uniform size, Ls, in the crystallizer at the saturation
temperature (Myerson, 2002). Process has to be conducted at low supersaturation lewel,
within the metastable zone width. Seeded crystallization process is influenced by the
way in which the seeds are produced, seeding policy, seed quantity and loadings
(Hojjati et al, 2005). The size and mass of seeds influences the crystal size distribution
and the crystal growth rate (Lung-Sommariba et al). Seeds must often be very small and
often are produced by milling. Seed size distribution has to be narrow. If the
supersaturation is maintained below the metastable zone upper limit, than the
spontaneous nucleation will be avoided and only the seed crystals will grow.
Consequently, produced crystal size distribution will be narrow. Some materials tend to
form agglomerates. In that case crystals with less uniform size distributions (bimodal)
will be obtained. Required seed mass can be calculated by the equation (Lung-
Sommariba et al):
3
m c ⎛ Lc ⎞ (1)
=⎜ ⎟
m s ⎜⎝ Ls ⎟⎠
where
mc and ms are the produced crystals and seed mass,
Lc and Ls are the final crystal and seed size.
The value of the Lc/Ls has to be between 1 and 5 (seed mass represents 2-5 % of the
final theoretical mass of crystals that can be obtained for the desired working
conditions).
1.2 Estimation of crystal growth kinetics
Garside and al., 1982, proposed a method for estimation of the growth kinetics from the
desupersaturation curve. The method assumes that no nucleation occurs, so the solution
concentration change is due to crystal growth only. If the secondary nucleation is
suppressed, kinetic parameters, the power g and constant kg, can be estimated from the
equations:
2 ⋅ F ⋅ ΔC 0 ΔC 0 ⋅ ΔC&&0 (2)
g= +
3⋅ ρ ⋅ L ⋅ A
0 T0ΔC& 0
− ΔC& 0 (3)
kg =
AT 0 ⋅ ΔC 0g
ΔC 0 - superaturation at time zero; ΔC& 0 , ΔC&&0 - first and second derivatives of the
desupersaturation curve at time zero; F – shape factor ratio (β/α); L0 - average seeds
size; AT0 – seeds surface area at time zero.
The derivatives ΔC& 0 , ΔC&&0 , are obtained by fitting the desupersaturation curve to a
polynomial of the form:
Δ C = a 0 + a 1 ⋅ t + a 2 ⋅t 2 (4)
So that ΔC 0 = a 0 , ΔC& 0 = a1 and ΔC&&0 = 2 ⋅ a 2 .
2. Experimental
Batch cooling crystallization of the pentaerythritol from the aqueous solution was
carring out in the Rushton dimensions crystallizer equipped with four baffles and axial
flow impeller (four pitched blade impeller). The size and the mass of the seed was
selected and calculated from the average crystal size obtained by unseeded experiment.
Finer fractions of industrially prepared pentaerythritol were used as seed. Defined mass
of seed was introduced to the crystallizer at the saturation temperature. All experiments
have been carried out at the same linear cooling rate (16 °C/hr). The solubility curve
and the metastable zone width were determined experimentally. Since the metastable
zone width is very narrow at higher mixing rates, mixing rate was set to 400 rpm.
Produced crystals were rinsed with cold acetone. Crystal size distribution was
determined by the sieve analysis and their shape by light microscope.
For estimation of the growth kinetics, from the desupersaturation curve during batch
seeded experiments, the Garside et al. method was used (equations 2 and 3). The crystal
growth is controlled by the surface integration (polynuclear or birth and spread
mechanism of grow) (Myerson, 2002), since the values of the exponent of growth, g, are
very high (g>2) (Table 1). Both parameters decrease with the increase of the seed size
and mass. The crystal growth rate is inversely proportional to seed size and loadings.
The influence of the seed mass and size on the crystal size distribution is shown on
figure 1. Only for the smallest seed size (90 μm) the unimodal narrow CSD is obtained.
Since the growth rate depends on the seed size additional nucleation (contact, breakage)
occurred. This additional nucleation is very intense, so large numbers of nuclei were
formed momentarily, resulting with broadening of the CSD with the increase of the seed
size. Unseeded experiments results with the widest CSD due to high level of
supersaturation and consequently very intense nucleation.
Figure 1. The influence of the seed size and mass on the CSD (τ=80 min)
Figure 2. The influence of the seed mass and the batch time on the CSD (Ls=90 μm)
In order to produce crystals with unimodal narrow size distribution, the batch time was
decreased from 80 to 35 min. Size and quantity of the produced crystals were smaller,
and unimodal CSD’s were obtained for all seeded experiments. Bimodal CSD was
obtained only for unseeded experiment, which means that secondary nucleation occurs
at short batch time also (process was also carried out within the metastable zone width).
The initial surface area of the seeds grows with the quantity of seeds added to the
solution (Table 1). Larger surface competes for grow, higher supersaturation was
achieved and consequently bigger crystals were produced. Modal crystal diameter is
pretty much the same for all experiments at the given batch time (~180 mm for 35 min,
and 550 for 80 min), but the mode fraction is highest for the smallest seeds.
The applicability of the LN and RRSB function for the CSD description was tested.
Since both functions can be used for description of the CSD (almost the same
correlation coefficient ~ 0,98), another criteria was used to select the appropriate
mathematical function. Based on the evaluated (from the LN and RRSB function
parameters) and experimental values of the specific surface area the adequate function
was selected (Table 2.). Since the evaluated values for the LN function differs highly
from the calculated values, RRSB function can be used for description.
Table 2. Specific surface area of the obtained crystals for two different experiments
duration
τ = 35 Sm, cm2/g τ = 80 Sm, cm2/g
min min
Ls, μm Calculated LN RRSB Ls, μm Calculated LN RRSB
Unseeded 200,42 846,15 164,62 Unseeded 150,23 572,38 110,40
90 275,61 2090,80 297,75 90 92,43 241,52 65,89
125 274,83 3114,45 330,24 125 102,56 284,89 72,87
180 251,02 1555,16 248,74 180 73,14 234,86 57,98
Longer batch time and bigger amount of the seeds, results with CSD shifted toward
coarser fractions (bigger x63) (Table 3). The shape of the crystals is not influenced by
the applied process condition (Figure3).
Figure.3. PET crystals taken on light microscope
4. Conclusions
More uniform CSD is obtained with shorter batch time and smaller amount of seeds
added in the just saturated solution. The seed size and loadings strongly influences the
crystallization kinetics parameters and the produced crystals size distribution. Maximal
value of supersaturation is achieved during unseeded crystallization. Its value decreases
with the increase of the average seed size, and the decrease of the amount of seeds.
Higher values of the size and amount of seeds results with lowering the growth kinetic
parameters (kg and g). High values of parameter g correspond to polynuclear or birth
and spread mechanism of growth. CSD can be approximated with the RRSB function.
Longer batch time and bigger seed size and mass shifts the CSD to the coarser fractions.
The shape of the crystals is not influenced by the applied process condition.
References
Chong K.L., Smith R., 2004, Optimization of batch cooling crystallization,
Chem.Eng.Sci, 59, 313-327.
Garside J, Gibilaro L.G, Tavare N.S., 1982, Evaluation of crystal growth kinetics from
desupersaturation curve using initial derivatives, Chem.Eng.Sci. 37, 1625-1628.
Genck W.J., 2003, Optimizing Crystallizer Scaleup, Chem. Eng. Prog. 99, 36-44.
Hojjati H., Rohani S., 2005, Cooling and seeding effect on supersaturation and final
crystal size distribution (CSD) of ammonium sulphate in a batch crystallizer, 44,
949-957.
Lung-Sommariba B.L.M., Moscosa-Santillan M., Porte C., Delacroix A., 2004, Effect
of seeded surface area on crystal size distribution in glycine batch cooling
crystallization: a seeding methodology, J.Crystal Growth, 270, 624-632.
Mersmann A., 1988, Design of Crystallizers, Chem. Eng. Process. 23, 213-228.
Myerson A.S., 2002, Handbook of Industrial Crystallization, Butterworth-Heinemann,
Boston.